Epileptic seizures don’t always last very long. And it’s not necessarily the seizure itself that does the harm – it can be the uncertainty. If you don’t know when one will strike, it can stop you from going for a swim, or driving your car. It can affect your employment – it can impact your whole life.
Seizures can, on the surface, appear to be fairly random. But one of our big research discoveries has been to pick up the long-term patterns of risk by looking over long timescales of data. In my work we go back into people’s seizure diaries over many years, then integrate information that we’ve gathered from their wearable devices. We can then look at combinations of factors that lead to periods of high risk.
We believe that for people with epilepsy, many biological rhythms can combine to create conditions that are conducive to seizures. It’s a bit like how multiple factors can combine to lead to very high risk of bushfires in Australia. We know there are patterns in our climate – seasonal changes, changes over many years, sudden shifts. All of those influences can combine to create conditions where extreme weather events are more likely.
It’s not necessarily the seizure itself that does the harm – it can be the uncertainty.
It’s the same thing with seizures. We can see this combination of different factors that lead to periods of high risk of seizures. We can now give people some advance indication of when they’re going into a high-risk phase of their cycle, And, just as importantly, a low-risk phase – so they can plan when they might be able to take a holiday, for example.
I have a user who’s been using our app who has a (roughly) 21-day cycle for his own seizure risk. He wrote to me to say, “I was on holiday with the grandchildren, and it was so stressful being with these little kids all the time. Luckily, I wasn’t in my high-risk state or I’m sure I would have had a seizure.”
It’s very gratifying to hear a patient use that knowledge to make plans.
My work is all about integrating data from multiple sources that record lots of different kinds of body measurements, and combining all of that information to help people understand their seizures.
Early research was done in a cohort of people who were implanted with a device on the surface of the brain, just underneath the skull. It’s quite invasive – that’s no longer happening. We’re trialling a new device (Epiminder) that’s less invasive; it sits between the bone and the skin and records brain activity. That’s a fantastic development, but it’s still in clinical trial stage.
You look at your risk factors in the morning the way people look at the weather forecast for the week ahead.
What we’re really excited about right now is this ability to infer people’s rhythms of seizure risk just by using a wearable device – so it’s a completely non-invasive signal – combined with their own record of their past seizure times. We’ve just been using a Fitbit, a consumer-grade smart watch. The signal we’re most interested in is heart rate, but there are a whole range of relevant signals: oxygen saturation, electrodermal activity – that’s the sweat sensor. Also skin temperature, sleep quality, and exercise levels.
The risk factors can be quite different for different people, so when you tie this data in with the individual’s recorded history, it becomes tailored for them. The “next big thing” really is patients just putting on their Fitbit, looking at the app on their phone, and getting as much data and insight as they were getting on any of the previous devices. You look at your risk factors in the morning the way people look at the weather forecast for the week ahead.
Of course, there are limitations. Capturing someone’s seizure history is much more accurate when we have the implant device, but that’s something that’s not available yet – and probably will never be available for everyone, especially in the developing world. Epilepsy is a big problem, and with a mobile and a wearable and remote cloud computing, we’re actually able to roll out this technology in a really widespread scalable way. That’s the power.
When you tie this data in with the individual’s recorded history, it becomes tailored for them.
If you’d asked me in high school what my dream job was, it was probably a science journalist writing for a magazine like Cosmos. I was always interested in medicine, biology, physics and maths, but I also loved English – writing was probably my favourite subject. But I was a shy kid and I think I realised how much interpersonal interaction there was in journalism. So I kept following the maths pathway in engineering, while retaining a really strong interest in medicine. Biomedical engineering was perfect.
I actually went through Melbourne University the very first year the degree was offered. “Medtech” might have just become a word, but there weren’t many medtech companies in Australia. The big name at that time was Cochlear. Beyond that, I didn’t know much about the industry. I had university classmates who ended up going into accounting firms, and that didn’t interest me. So I did a PhD just to stay in the field that I love.
By the time I finished my PhD, there were so many more start-ups in the medtech area, doing all sorts of wonderful things with mobile devices and apps and wearables. I became one of the team members of Seer, which specialises in epilepsy management, from the diagnostic stage through to the lifelong management journey. They’ve developed what we call clinical-grade wearables, which means devices that record signals at the same level you would expect from a hospital-grade device but can be done in the home. Today, this includes consumer wearables like the Fitbit, which I use in my research and with the Seer app. We also collaborate with companies like Epiminder, who are doing implantable devices in the brain. It’s the whole scope.
When I can, I love to get into the code.
I do less of the actual data crunching and programming than I would like now. The further you get through your research career, the more it becomes about the funding applications and the high-level ideas and strategy, so you pass over the number crunching to PhD students and other research assistants. And I have a lot of support from our industry partner, Seer – they have software developers and user experience researchers who are really helping to speed up this clinical translation. But when I can, I love to get into the code.
The most exciting part of my work these days is finding new results from the data. We’ve discovered that there are these really interesting long-term rhythms in heart rate that appear to affect people’s seizure likelihood. By the time those results had been published, we were already piloting that signal in a forecast in the mobile app with the same participants. That kind of turnaround is unbelievable in terms of translating academic research into currently available technology. It’s sort of unheard of.
As told to Graem Sims for Cosmos Weekly.